Automated human sperm tracking using mean shift - collision detection and modified covariance matrix method

摘要

In vitro fertilisation (IVF) is a popular technique in assisted reproductive technology. The success of IVF mainly depends on the selection of the correct sperm in human semen sample. Sperm tracking plays an important role in selecting the active-moving sperm. One of the major challenges in sperm tracking is the collision of sperm cases during tracking. To solve this issue, mean shift-collision detection and modified covariance matrix (MS-CDMCM) is proposed. Specifically, MS-CDMCM detects collision and generates a new covariance matrix based on the collision condition. Then, this new covariance matrix will form a new tracked region to continue the tracking process. Results show that the proposed method is a more accurate and robust tracking method than other state-of-the-art sperm tracking methods. The proposed method produces significantly low error values, such as MAE, MSE and RMSE, according to the quantitative analysis when compared with ground truth images. The proposed method is expected to be implemented in sperm motility assessment in the future.